Sunday, August 25, 2013

ReachTEL (State): The Same Story Told Many Times

ReachTEL (Tas State): Lib 51.1 ALP 24.4 Green 14.7 Other 4.2 Undecided 5.6
Interpretation: Lib 53 ALP 27 Green 16 Other 4
Outcome based on this poll "if election was held now": Liberal Majority Win (14-7-4 or 15-7-3)
Projection based on all polling: 14-7-4

This article concerns Tasmanian state polling.  For the Tasmanian federal poll from the same polling run see ReachTEL Says Tas Labor Still Losing Three.

A new State ReachTEL is out and it's another sad story for the Labor-Green minority government's re-election prospects, or even for anyone who believes that the prospect of a close election encourages accountability by both sides to the electorate.  It's a very happy story, of course, if you're a diehard Tasmanian Liberal. 

It's the same story that has been told in ten EMRS polls and three previous ReachTELs since the start of 2011 - that the Liberal Opposition is headed for majority government.  Every poll has projected to at least thirteen seat wins for the Liberals, and most have projected to more.  Each has told a slightly different story about the potential scale of the victory and the possible division of the Government's seat losses between Labor and the Greens. This one sits pretty much in the middle of the fairway.  The overall picture is relentlessly consistent, and the clock is ticking down with only seven months to go.  Even a new Abbott federal government will have their work cut out to do something annoying enough for the Liberal Party to stuff this one up from here.

This ReachTEL shows a slight improvement in Labor's position since the first public state ReachTEL in June (State Labor Back To The Lifeboats).  Compared to that poll, on the headline rate Labor has recovered by 4.7 points, and all other categories are down:  Liberals by 1.4, Greens by 0.9, Other by 1.8 and Undecided by 0.6.  The increase in Labor's share of the vote is strongly statistically significant, to the point of being more or less certainly real, but the baseline for the comparison was the worst poll of the government's term, so there's not a lot of consolation there.

The three northern seats were also polled in July (The North Remembers).  The average polled change in those seats since that poll after removing undecided voters is Labor +1.5, Liberal -0.7, Green -1.9, Ind/Other +1.  The biggest recovery in Labor's stocks since the June poll has been in Bass and Braddon.  In Bass it has come at the expense of the Greens, and in Braddon the very high Ind/Others rate recorded in the June poll did not recur in the July or August polls.

Here are the figures from the current poll, with undecided voters excluded, translated into likely quotas and seat outcomes.  I have not applied any house effect corrections to these figures.

A small note of caution is warranted about the Green figure in Braddon.  The Green vote for Braddon in the federal poll (about four points) seemed unrealistically low so the sample as a whole might be less Green-friendly than average for the seat.  However it would have to be wrong by about 3 points to make the Greens competitive in the seat, and a Green state vote of 8 in Braddon doesn't seem unrealistic by itself.

The individual projections from this sample are mostly straightforward.  The one that is tricky is Bass, which sits on the cusp between 4-1-0 and 3-1-1 in this sample.  It is unclear because there is a high chance that the Liberals, based on leakage and Labor preferences, would lose votes to Kim Booth as the cutup went on, so that he would be closer to one quota after a while than they were to four.  But even assuming Booth reached the Liberals' primary figure, he would still be at the mercy of, you may have guessed it, the dreaded Ginninderra Effect.  For those who haven't read that link already (or more likely found it way over their heads) the Ginninderra occurs when a lone candidate is nearer to their quota target than a rival party is to its but he/she has fewer votes than each individual candidate from that group.  The party that appears closer on raw totals (say, 0.7 vs 3.6) can lose if its lone candidate is stuck behind the individual totals of each candidate from the rival party and eliminated.  Nasty stuff.

The reason I haven't applied any house effect correction is that even if ReachTEL polls do have a small pro-Coalition house-effect (on which the federal evidence is lately less clear than it used to look) it really doesn't affect the above picture much.  It would still be unknowable from the figures whether or not the Greens might lose Bass in the manner mentioned above.

On the whole the Liberals have a comfortable 14 seats according to this poll, and the Franklin breakdown is projecting defeat for either Lara Giddings or David O'Byrne. Restoration of a 35-seat House now seems to be completely off the table, but for what it's worth it would result in about 19-21 Liberal seats, 9-10 Labor and 4-5 Green.

There are also some figures on "the performance of the current Tasmanian state government".  This was also polled in June and the results are incredibly dire:

Despite the improvement in Labor's state vote from the June poll, there is no improvement in what voters think of the state government, which has an appalling net performance rating of -44.7.  An improvement in the government's standing in Braddon matches an 11 point rise in Labor's polled vote there since the June poll, but in Denison and Lyons the government has lost favour without losing votes.  It is not as if we see a well-regarded Government outshone by a stellar Opposition.  Rather, we see a government viewed as so disappointing that its disapproval rate exceeds even the Liberal vote by 8%. 

For some time I have been attempting a three-party aggregate of Tasmanian state polling data, in which the old data is downweighted with each new poll included.  The aggregate assumes that there won't be any independents or other parties winning seats, until such time as polling evidence emerges that someone has a serious chance.  Votes for Ind/Others are redistributed equally between the three parties.  I've made the new poll data 40% of the new aggregate and the old data 60%.  This is what it looks like after that:

Although the Liberals have three quotas in Franklin in my aggregate, they would suffer from leakage from Will Hodgman's surplus, and the high profiles of Giddings and O'Byrne would assist Labor in obtaining leakage from other parties.  So it is not clear where that seat would go. 

Although the individual seat aggregates give the Greens four seats, two of them are with only just enough vote share to be confident that they would win the seat with that exact share.   The Labor seat in Braddon is also a bit shaky.  To try to work out where to assign the seats for my projection on the front page, I assigned rough probabilities for each seat based on the idea that the aggregate would be slightly out in various directions in each seat.  When I added these up, the totals came to 14.3 for the Liberals, 7.3 for Labor and 3.4 for the Greens.  The Greens therefore are allocated the last seat by a whisker and my aggregate goes from 14-8-3 to 14-7-4, which would be a mirror image of the thrashings handed out by Labor under Jim Bacon in 2002 and Paul Lennon in 2006. 14-8-3 and 15-7-3 are the two other high-probability outcomes, but 15-7-3 relies on the Liberals bagging a four somewhere, which is always going to be easier said by the model than actually done.

In the normal course of things an EMRS poll would be expected around this time.

PS:  There have been so many polls saying the same thing now that they no longer seem to invoke too many exotic reactions.  ALP state secretary John Dowling said "The ReachTEL data, compared to the data I have seen, overestimates the Liberal vote by 3-4 per cent".  It is difficult to benchmark ReachTEL (or indeed, any) polls for Tasmanian state elections since the only regular comparison data comes from EMRS, and EMRS have been far from accurate at past state elections.  However, even taking 3.5 points from the Liberals in every electorate still results in them easily winning majority government with 13 seats. 

The Sunday Tasmanian today published statewide age and gender results.  They show a gender divide similar to that being commonly found in national polls, only more so.  The age results are unremarkable, except that the Greens do not do significantly better in the under-35 bracket than in the 35-50 and 51-65 age brackets (Labor does.)  Females (27.1% ALP, 46.1% Liberal, 16.3% Green. 2.5% Other, 7% Undecided) are about ten points less hostile to the Government than males (ALP 21.6%, Lib 56.6, Green 12.9, Other 4.9, Undecided 4).  What the statewide gender figures hide is that this gender divide effect is strongly concentrated in the electorates where the Liberal Party is performing best, especially Lyons and Braddon.  (In Bass males buck the state trend by being more likely to vote Green than females, which seems odd and is possibly sample error noise rather than a real result.)

The ABC here writes "A swing against the Greens in Braddon saw Labor's primary vote in that electorate jump 11 per cent, with the Liberals gaining only 1 per cent."  However the "swing" against the Greens is only 1.3% on the raw figures, which is not even statistically significant.  The real "swing" in Braddon since the June poll was in Other (down from a suspiciously high 8.4 points to 2.3) and Undecided (down from 7.4 to 3).  With Undecided removed, Labor is up 10.4 points in Braddon, at the expense of the Liberals (1.7), Greens (1.8) and Other (6.6).

No comments:

Post a Comment

The comment system is unreliable. If you cannot submit comments you can email me a comment (via email link in profile) - email must be entitled: Comment for publication, followed by the name of the article you wish to comment on. Comments are accepted in full or not at all. Comments will be published under the name the email is sent from unless an alias is clearly requested and stated. If you submit a comment which is not accepted within a few days you can also email me and I will check if it has been received.